Lower semicontinuity via W^{1,q}-quasiconvexity

LOWER SEMICONTINUITY VIA W 1 «-QUASICONVEXITY

JEAN-PHILIPPE MANDALLENA

Abstract. We isolate a general condition, that we call "localization prin-
ciple", on the integrand L : M — > [0, oo], assumed to be continuous, un-
der which W 1,9 -quasiconvexity with q £ [l,oo] is a sufficient condition for
I(u) = L(Wu(x))dx to be sequentially weakly lower semicontinuous on
W 1,p (£l; M m ) with p S]l,oo[. We show that this "localization principle" is
satisfied under hypotheses on L which are related to the concept of fast growth
integrand introduced by Sychev.

1. Introduction

Let m, N > 1 be two integers, let f2 C M. N be a bounded open set with Lipschitz
boundary, let M := M mxN , where M mxAr denotes the space of all real m x N
matrices. Let p £]l,oo[, let L : M — > [0, oo] be a continuous function and let
I : W^ P {VL- W l ) -> [0, oo] be defined by

I{u) := ( L{Vu(x))dx.
Jn

In [BM84 Ball and Murat introduced the concept of M /1 ' 9 -quasiconvexity for q £
[1, oo], i.e., L is W 1,9 -quasiconvex if and only if

J L (Vu(y)) dy > L(f) for all u £ Z € + Wq' 9 (Y; W n )

with lt(y) := £y and Y :=] - §, \[ N , and proved (see [BM84| Corollary 3.2]) that
J^' p -quasiconvexity is a necessary condition for / to be sequentially weakly lower
semicontinuous (swlsc) on W 1,p (ft;R m ), i.e., when

Un u in W 1,p (Q;R m ) implies lim I(u n ) > I(u).

However, proving that iy 1,p -quasiconvexity, or some variant of it, is also sufficient is
still an open problem. In this paper we isolate a general condition on L (see (C Pj9 )
in Theorem I l.lj) under which M^ 1:9 -quasiconvexity is a sufficient condition for I to
be swlsc on W 1,p {Vl\ R m ). More precisely, our first main result is the following.

Theorem 1.1. Given p £]l,oo[ and q £ [l,oo], assume that L is W 1,q -quasiconvex
and satisfies

(C p ,g) for every £ £ M and every {v n } n C W^ p (Y;M. m ) such that

( v n ^l ( in W^ p (Y;W n );

| sup / L(Vv n (y))dy < oo,
K n Jy

Key words and phrases. Weak lower semicontinuity, W 1,9 -quasiconvexity, Young measures,
cqui-integrability.

2

JEAN-PHILIPPE MANDALLENA

there exist a subsequence {v n }„ (not relabeled) and {w n }n C l( +W ' <1 (Y; M. m )
such that

J |Vw n — Vw n \ — > in measure;
\ {L(\7w n )} n is equi-integrable.

Then, I is swlsc on W^(ft;]R m ).

As a direct consequence of Theorem 11.11 we have

Corollary 1.2. Given p e]1,oo[, if (C PjP ) holds then W 1 ' p -quasiconvexity is a
necessary and sufficient condition for I to be swlsc on W ,p (fi;R m ).

In fact, Acerbi and Fusco (see |AF84] ) showed that W 1,00 -quasiconvexity is suffi-
cient for / to be swlsc on W 1,p (f2;R m ) provided that L has p-growth, i.e., L(-) <
a(l + | • | p ) for some a > 0. We remark that the key argument in their proof is in
fact the following result, which we call "localization principle":

(A) for every ^eM and every {v n } n C W 1 ' P (Y; R m ) such that

Vn^k in W lj, {Y;R m ),

there exist a subsequence {v n } n (not relabeled) and {w n } n C l^+C^°(Y; R m )
such that

J |V«rj — Vw„| — > in measure
[ {|Vw n | p } ra is equi-integrable.

Note that (A) is a particular case of the decomposition lemma (for more details see
Kristensen |Kri94j and also Fonseca, Muller and Pedregal IFMP98] ). Using this "lo-
calization principle" Kinderlehrer and Pedregal (see |KP92j and also |Syc99| ) proved
Acerbi-Fusco's theorem by using Young measure theory. Kinderlehrer-Pedregal's
approach was extended by Sychev (see |Syc05| ) to the case where L has fast growth,
i.e., (3G(\ ■ |) < £(•) < a(l + G(\ ■ )) for some a,/3 > and some convex function
G : [0, oo[— > [0, oo[ such that lim^oc tG'(t)/G(t) = oo and tG'(t)/t is increasing for
large t. We also remark that the key argument in its proof is still a "localization
principle", more general than (A), i.e.,

(B) for every £ e M and every {v n } n C W 1 ' P (Y; M. m ) such that

J v n l{ in W 1 ' p (Y;R m )
j sup J Q G(\Vv n (x)\)dx < oo,

there exist a subsequence {v n } n (not relabeled) and {»„}„ C l^+C^°(Y; R m )
such that

J |Vu n — Vw„| — > in measure
^ {G(\Vw n \)} n is equi-integrable.

It is easily seen that (C Pi9 ) generalises (A) and (B) in a natural way, i.e.,

J if L has p-growth then (A) implies (C Pi00 )
1 if L has fast growth then (B) implies (C Pj00 ),

which makes that Theorem 11.11 contains Acerbi-Fusco's theorem and Sychev's theo-
rem in the homogeneous case. (Note that the validity of (Cj, )00 ) implies the validity
of (C Py q) for all q G [1, oo].)

According to Theorem 11.11 it is natural to try to find new conditions on L (dif-
ferent from p-growth and fast growth) under which the "localization principle" is
satisfied. In this paper, inspired by the work of Sychev (see |Syc0 5 ), we introduce
conditions on L (see (D1-D3) in Theorem ll.3|) under which (C PiP ) holds for all

LOWER SEMICONTINUITY VIA W 1,<! -QUASICONVEXITY

3

p G]N,oo[. (Note that these conditions are satisfied when L has fast Growth, see
|Syc05[ Proposition 3.1 and Lemmas 3.3 and 3.4].) More precisely, our second main
result is the following.

Theorem 1.3. Assume that L is finite and satisfies:

(Di) there exists A :]0, oo[— >]0, 1[ such that A(i?) — > 1 as R — ¥ oo and

hm sup — — — — = 0;

R -^°° \t\>R Ms)

(D2) there exists u\ > such that

£(«0 <ai(l+i(0)
for all £ e M and all t e]0, 1[;
(D3) i/iere exist ct2 > and e > such that for every (el and every t £ [0, 1[
there exists c^j > for which

L« + «C + e)<a 2 (l+L(£ + C))
/or a// s € [0,t], a// C G M wit/i |C| > c^ t and all e € M wi£/i |e| < e.
T/ien (Cp, p ) ZioMs /or all p G]N, oof.

The plan of the paper is as follows. Theorem 1 1.1 1 is proved in Section 3. Its proof
uses some classical facts on Young measures that we recall in Section 2. (Note that
it seems to be difficult to prove Theorem 11.11 without using Young measure theory.)
Finally, Theorem ll.3l is proved in Section 4.

2. Some facts on Young measures

Young measures were introduced by Young in 1937 (see |You37j ) with the purpose
of finding an extension of the class of Sobolev functions for which one-dimensional
nonconvex variational problems become solvable. In the context of the multidimen-
sional calculus of variations, Kinderleherer and Pedregal (see [KP921 [KP94] ') and
independently Kristensen (see |Kri94j ) were the first to use Young measures for
dealing with lower semicontinuity problems. Relaxation and convergence in energy
problems were studied for the first time by Sychev via Young measures following a
new approach to Young measures that he introduced in |Syc 99 . In this section we
only recall the ingredients that we need for proving Theorem ll.il For more details
on Young measure theory and its applications to the calculus of variations we refer
to |Ped97llP^d00l|Syc04| .

Let V(M) be the set of all probability measures on M, let C(M) be the space of all
continuous functions from M to 1 and let

C (M) := ($ € C(M) : lim = oj.

L |£|-K> J

Ifl-

Here is the definition of a Young measure.

Definition 2.1. A family (f^ x )xeQ of probability measures on M, i.e., fj, x 6 P(M)
for all x S ft, is said to be a Young measure if there exists a sequence {£n}ri of
measurable functions from f2 to M such that

$(&0 A A*(-)> in for all $ e C (M)

with ($; ^(.)) := / M In this case, we say that {£„}„ generates (^x)xeo

as a Young measure.

4

JEAN-PHILIPPE MANDALLENA

The following lemma makes clear the link between convergence in measure and
Young measures. (The proof follows from the definition.)

Lemma 2.2. let and {Cn}n be two sequences of measurable functions from

to M. If generates a Young measure and if |£ n — £ n | in measure then

{Cn}n generates the same Young measure.

The following theorem gives a sufficient condition for proving the existence of Young
measures (for a proof see [Bal89l |Syc04} IFL07] k

Theorem 2.3. Let 9 : M —> R be a continuous function such that lim^i^oo 9(Q =
oo and let {£n}n be a sequence of measurable functions from to M such that

sup / 6(£ n (x))dx < oo.
n Jn

Then, {£, n }n contains a subsequence generating a Young measure.

The following two theorems are important in dealing with integral functionals (for
proofs see [Bal84| [Syc99] ).

Theorem 2.4 (semicontinuity theorem). Let L : M — > [0,oo] be a continuous
function and let {£n}n be a sequence of measurable functions from D, to ML such
that {£ n }n generates (fJ. x )xen as a Young measure. Then

lim / L(£ n (x))dx> / (L;fj, x )dx.
n->oo Jn Jn

Theorem 2.5 (continuity theorem). Let L : M — > [0, oo] be a continuous function
and let {£ n }n be a sequence of measurable functions from Q to M such that {£ n }n
generates (nx)xen as a Young measure. Then

lim / L{^ n {x))dx = / (L;fi x )dx < oo
n ^°°Jn Jn

if and only if {£(£„)}„ is equi-integrable.

3. Proof of Theorem 11.11

Let {u n } n C W 1 '*'(fi;M m ) and let u G W 1 ' p (0,;R m ) be such that u n ->• u in
W 1 ' p (n;M m ). We have to prove that

(3.1) lim I(u„) > I(u).

n— >oo

Step 1: localization. Without loss of generality we can assume that:

(3.2) ||«„ - u||z,p(n ; R"») 0;

(3.3) oo > lim I{u n ) = lim I(u n ) and so sup / L(\7u n (x))dx < oo.
As u n u in W rl >P(0;R TO ) we have

(3.4) sup / \\7u n (x)\ p dx < oo,

n Jn

and so, by Theorem 12. 3[ there exists a family (fi x )xeO of probability measures on
M such that (up to a subsequence)

(3.5) {Vu n }n generates (p x )xen as a Young measure.

LOWER SEMICONTINUITY VIA W rl ' a -QUASICONVEXITY

5

From Theorem 12.41 it follows that

lim I(u n ) > / (L: fi x )dx

n— >oo JO

with (because (|3.3I) holds) for a.e. xo G fi,

(3.6) (i;Mx ) < °°.

Thus, to prove (|3.ip it is sufficient to show that for a.e. xo G f2,

(3.7) (L;» X0 ) >L(Vu(x )).

Step 2: blow up. From (|3.3I) we deduce that there exist / G i 1 (^; [0, oo[) and
a finite positive Radon measure A on SI with |supp(A)| = such that (up to a
subsequence) L(SJ u n )dx fdx + A in the sense of measures and for a.e. x$ G O,

(3.8) lim lim f L(Vu n (x))dx = f(xo) < oo

r^On^ooJ Xo+rY

with Y :=} — |, |[ . By the same argument, from (I3.4[) we see that

(3.9) lim lim / \Vu n (x)\ p dx < oo.

As u G PF 1 ' p (il;R m ) it follows that u is a.e. LP-differentiable (see |Zie891 Theorem
3.4.2 p.129]), i.e., for a.e. x Q G Q.,

1 up
( 3 - 10 ) ^To ^v^P( x o + ') - «(»o) - Vu(x )2/|| iP(ry . Rm) = 0.

From (I3.2[) we see that (up to a subsequence) for a.e. xo G fi,

(3.11) |u„(z )-u( : ro)| p 0.

As C (M) is separable we can assert that for a.e. Xq G fl, x$ is a Lebesgue point of
<*5At(.)> for a11 * G C (M), i.e.,

(3.12) lim/ (<&,Hx)dx = {$,/i xo ) for all $ G C (M).

T ^°J xo+rY

Fix any i eO such that (|3l| , (pO]) . ([3TT0]) . ((3~TTj) and (|3~T2|) hold and fix r >
such that xo + rY C £1 for all r G]0,ro]. For each n > 1 and each r G]0, ro], let
ujj G W /1,P (F; M m ) and a family (Hy) y eY of probability measures on M be given by

<(y) : = J- ("nf^o + ry) - u n (x ))
Then (J3T5J) (resp. t\'3.9]i ) can be rewritten as

(3.13) lim lim / L(X7 u r n {x))dx < oo (resp. lim lim / \\7u 7 n (x)\ p dx < oo).

r— >-0 n— >oo Jy r— >0 n— >oo Jy

Taking Q3.5P into account it is easy to see that for every r G]0, r ], {V«J|} n generates
(Py) y( zY as a Young measure, i.e.,

(3.14) *(V<) A ($, in L°°(Y) as n -> oo for all $ G C (M),
and using p. 121) it is clear that

(3.15) ($; A jUx ) in as r -> for all $ G C (M).

6

JEAN-PHILIPPE MANDALLENA

On the other hand, we have

\Wn.r -hu(x )\\ V L p(Y^) = j Y \ u n,r{y) - hu(x„)(y)\ P dy

= r N+p ^ U n{ x + ') — u n{%o) — 'vu(io) Hz,P(rY;R m ) '

and consequently

i 1 1 n C

+^v+F ll^o + O - "fro) - Z v «(*o) HiP (rr;Rm)
with c > which only depends on p. Using (|3.2|) . p. lip and (|3.10p we deduce that
(3.16) lim lim \\u r n - l Vu i Xo ) \\lp(y-s.™) = 0.

According to (|3.16[) . (|3.13p and (|3.14p together with (|3.15p . by diagonalization there
exists a mapping n — > r n decreasing to such that

IK™ - hu(xo) \\lp(Y;R'™) ~>

lim f v \\7u'L rl (y)\ p dy < oo, and so sup f v | Vi£ n (y)\ p dy < oo

n— >oo n

lim f v L(VuI> {y))dy < oo, and so sup f v L(V 'ul> (y))dy < oo

, {V«^} n generates fi Xo as a Young measure,

and consequently we have:

f « n -Z Vu(xo) m^(F;R-)
( d - 17 ) < sup/ y L(Vv„(j/))<fy <oo;

I n

(3.18) {Vw„}„ generates /x^ as a Young measure,

where w n := m!>.

<

Step 3: using (C Piq ) and VF^-quasiconvexity. According to (|3.17p . by (C P 9 )

Oo)

there exists {w„}„ C ivti(i D ) + W ,q (Y;M. m ) such that

|V«„ — Vio n | — >• in measure
L(Vw n ) is equi-integrable,

hence, by (|3.18p and Lemma [2121 {Vw„}„ generates /it^ as a Young measure, and,
taking (|3.6p into account, from Theorem 12.51 we deduce that

(3.19) lim / L(Vw n (y))dy = (L;n X0 ).

As L is iy 1,9 -quasiconvex, we have

L(Vw n (y))dy > L(Vu(x )) for all n > 1,
and (|3.7p follows by letting n — >■ oo and using (|3.19p . ■

Y

Remark 3.1. In case q = oo the condition of W^'^-quasiconvexity is the classical
condition of quasiconvexity by Morrey (see |Mor52j ) .

LOWER SEMICONTINUITY VIA W 1,<! -QUASICONVEXITY

7

Remark 3.2. In fact, we have also proved that if {u n } n C W 1,p (Sl; R m ) is such that
sup,j Jq L(Vu n (x))dx < oo and if {Vu n } n generates (fi x )xen as a Young measure,
then for a.e. x £ Cl, fi x is a homogeneous gradient L- Young measure centered at
Vu{x), with u £ W lj >(il;R m ), provided that u n u in W l ' p {Vt;W n ) and {C p . q )
holds with q £ [1, oo]. Homogeneous gradient L- Young measures were introduced
and completely characterized by Sychev in |SycOO| where we refer the reader for
more details.

Remark 3.3. From the proof of Theorem 11.11 we can extract the following lower
semicontinuity theorem with the biting weak convergence.

Theorem 3.4. Given p £]l,oo[ and q £ [1, oo], assume that L is W l ' q -quasiconvex
and satisfies (C Pl ,). Then, for each {u n } n £ W^ifl; R m ) and each u £ W 1 '"^; R m )
such that u n u in W ' p (p.\ R m ) and sup n L(X7u n (x))dx < oo, there exists a
subsequence {u n } n (not relabeled) and a family (fi x ) x ^n of probability measures on
M such that:

(i) {Vu n } n generates (fi x ) x ^n as a Young measure;

(ii) L(\Ju n ) (L; /!(•)), where "— denotes the biting weak convergence;
(hi) (L; fi x ) > L(\7u(x)) for a. a. x £ f2.

For a deeper discussion of weak lower semicontinuity in the sense of biting lemma,
see Ball and Zhang |BZ90] (see also |Syc05| Lemma 3.2] for a simple proof of the
biting lemma).

4. Proof of Theorem 11.31
Let p £]iV,oo[, let £ £ M and let {v n } n C W 1 ' p (Y;R m ) be such that:

(4.1) v n It in W 1 *(y;R m ) ) and so sup f \Vu n (y)\ p dy < oo;

n Jy

(4.2) sup / L{Vv n {y))dy < oo.

n JY

As p > N, (|4.ip implies that, up to a subsequence,

(4.3) \\Vn-kh°>>(Y^)^0-

Step 1: using the biting Lemma. First, recall Sychev's version of the biting
lemma (see |Syc05| Lemma 3.2]).

Lemma 4.1. Let {/„}„ £ L 1 (Y; [0, oof) be such that sup„ f Y f n {y)dy < oo. Then,
there exist a subsequence {f n }n (not relabeled) and {M n } n c]0,oo[ with M n — > oo
such that {fnXY^yn is equi-integrable with XY n denoting the characteristic function
ofY n :={yeY n :f n (y)<M n }.

Taking (|4.2[) into account, from Lemma I4TT1 that we apply with /„ = L(Vv n ), we
can assert that, up to a subsequence,

(4.4) {L{yv n )xY n }n is equi-integrable.

Let {R n }n c]0, oof be given by R n := ess mf y £ y\Y n \^ v n(y)\- As L is finite and
continuous, Y \ Y n = {y £ Y : L(Vv n (y)) > M n } and M n — > oo, we have R n — > oo.
Let {u n }n £ W^iYiM™) be defined by

u n ■= Kv n with X n := A(i?„),

8

JEAN-PHILIPPE MANDALLENA

where A :]0, oc[->]0, 1[, with X(R n ) -»■ 1, is given by (Di). From ([43]) and we
have:

(4.5) ||V«„ - Vu„|| LP( y ;Mm xiv) -)• 0;

(4.6) ||«n-ifi|Uoo ( y^m)->0.

On the other hand, given any n > 1, L(Vu n ) = L(A n Vu n )xr ra + L(A„Vw„)xy\y ii ,
andsoL(Vu„) < ai(l+L(V«„)xyJ+£(A„Vu„.)xr\Y„ by using (D 2 ). But |Vu„| >
R n on y £ y \ Y~„ , hence

L(Vu„) < a x (1 + L(Vw„) X yJ + sup ^^L{\7v n ).

|f|>-R„ Hi,)

Taking (|4. 2[) and (I4.4[) into account and noticing that sup| ? | >i?ii — >• by (Di)

we conclude that

(4.7) {L(Vw„)}„ is equi-intcgrable.

Step 2: cut-off method. Given any n > 1, set:

c„ := c ?: a„ with c S)An > given by (D 3 );

6 n := sup L{Q with e > given by (D 3 ).

ICI<l?l+c„+e

(Such a exists because L is finite and continuous. Moreover, up to a subsequence,
either 9 n — > oo or {9 n } n is bounded.) Let <f> n 6 C£°(Y; [0, 1]) be a cut-off function
between Q n :=] e "~ 1 , ~ 2 S " and V such that ||</>n||z,=°(y) < ^- with

1 I

En : = -7r\\ u n — hWh^fY-m^)-
"n

(Note that, by gU), e„ -> and 6 n e n -> 0.) Define w„e! ? + W^'^F; M m ) by

:= If + A„0„(u„ - ^)
(with A» G]0, 1[ and A„ -)• 1). Then

Vm„ on Q„

£ + A„0„(Vw„ - + ® ( u n ~ k) ° n ^ \ <5»-

Setting C„ := {y € Y \ Q„ : \Vu n (y) - £| < c n } we have

(4.8) L{Vw n ) < L(Vm„) + L{Vw n ) X c n + L{Vw n )x Y \{Q n uc n y

But |£ + \ n <j> n (Vu n - £) + A„V</>„ ® (u n - lf)| < |£| + c„ + 2e„ on C n and e n -> 0,
hence |£ + A„^n(Vu„ - £) + A„V</>„ (g> (u n - lf)| < |£| + c n + e on C n , and so

(4-9) L(Vw„)xc„ <0nXY\Q n -

Moreover, as \X n (j)n\ < A n , |Vu n -£| > c„ on Y\(Q„UC n ) and |A„V0„® (tt„-Zf)| <
2e„ — ► 0, from (D 3 ) we see that

(4-10) L(Vw n ) X Y\(Q n uc n ) < aa(l + £(Vu»)).

Combining (|4.9[) and (|4.10[) with (|4.8[) we deduce that for every n > 1,

i(V%) < »„Xy\Q„ + (or a + 1)(1 + £(Vn„)).

Taking (|4.7p into account and noticing that 9 n \Y \ Q n \ = (N + o(l))6 n e n — > 0, we
deduce that

{L(Vw„)}„ is equi-integrable.

Vio„

LOWER SEMICONTINUITY VIA W 1,<! -QUASICONVEXITY

9

On the other hand, it is easy to see that there exists K > 0, which only depends
on p, such that

\Vw n - Vtinl* < K (|£| p + \Vv n n ((1 - X n f + X Y\ Qn ) + Kel.

Taking (14. ip into account and recalling that \ n — > 1, \Y \ Q n \ —> and e n — > 0, we
deduce that ||Vio n — Vu n \\ L p( Y :M mxN ) ~^ 0, and so ||Vi;„ — 'Vw n \\x / pfY;M m >< N ) ~ ►
by combining with (|4. 5[) . It follows that

\S7v n — Vw n \ — ¥ in measure,

and the proof is complete. I

Acknowledgments. I gratefully acknowledges the many comments of M. A. Sy-
chev during the preparation of this paper, and for the lectures on "Young measures
and weak convergence theory" that he gave at the University of Nimes during may-
june 2011.

The author also wishes to thank the "Region Languedoc Roussillon" for financial
support through its program "d'accueil de personnalites etrangeres" which allowed
to welcome M. A. Sychev, from the Sobolev Institute for Mathematics in Russia,
in the Laboratory MIPA of the University of Nimes.

References

[AF84] Emilio Acerbi and Nicola Fusco. Semicontinuity problems in the calculus of variations.

Arch. Rational Mech. Anal, 86 (2): 125-145, 1984.
[Bal84] E. J. Balder. A general approach to lower semicontinuity and lower closure in optimal

control theory. SIAM J. Control Optim., 22(4):570-598, 1984.
[Bal89] J. M. Ball. A version of the fundamental theorem for Young measures. In PDEs and

continuum models of phase transitions (Nice, 1988), volume 344 of Lecture Notes in

Phys., pages 207-215. Springer, Berlin, 1989.
[BM84] J. M. Ball and F. Murat. VK 1,p -quasiconvexity and variational problems for multiple

integrals. J. Fund. Anal, 58(3):225-253, 1984.
[BZ90] J. M. Ball and K.-W. Zhang. Lower semicontinuity of multiple integrals and the biting

lemma. Proc. Roy. Soc. Edinburgh Sect. A, 114(3-4):367-379, 1990.
[FL07] Irene Fonseca and Giovanni Leoni. Modern methods in the calculus of variations: L p

spaces. Springer Monographs in Mathematics. Springer, New York, 2007.
[FMP98] Irene Fonseca, Stefan Muller, and Pablo Pedregal. Analysis of concentration and oscil-
lation effects generated by gradients. SIAM J. Math. Anal, 29(3):736-756 (electronic),

1998.

[KP92] David Kinderlehrer and Pablo Pedregal. Weak convergence of integrands and the Young
measure representation. SIAM J. Math. Anal, 23(1):1-19, 1992.

[KP94] David Kinderlehrer and Pablo Pedregal. Gradient Young measures generated by se-
quences in Sobolev spaces. J. Geom. Anal, 4(l):59-90, 1994.

[Kri94] Jan Kristcnscn. Finite Junctionals and Young measures generated by gradients of
Sobolev functions. PhD thesis, Technical University of Denmark, Kyngby, 1994.

[Mor52] Charles B. Morrey, Jr. Quasi-convexity and the lower semicontinuity of multiple inte-
grals. Pacific J. Math., 2:25-53, 1952.

[Pcd97] Pablo Pedregal. Parametrized measures and variational principles. Progress in Nonlin-
ear Differential Equations and their Applications, 30. Birkhiiuser Verlag, Basel, 1997.

[PedOO] Pablo Pedregal. Variational methods in nonlinear elasticity. Society for Industrial and
Applied Mathematics (SIAM), Philadelphia, PA, 2000.

[Syc99] M. A. Sychev. A new approach to Young measure theory, relaxation and convergence
in energy. Ann. Inst. H. Poincare Anal. Non Lineaire, 16(6):773-812, 1999.

[SycOO] M. A. Sychev. Characterization of homogeneous gradient Young measures in case of
arbitrary integrands. Ann. Scuola Norm. Sup. Pisa CI. Sci. (4), 29(3):531-548, 2000.

10

JEAN-PHILIPPE MANDALLENA

[Syc04] M. A. Sychcv. Young measures as measurable functions and their applications to varia-
tional problems. Zap. Nauchn. Sem. S.-Peterburg. Otdel. Mat. Inst. Steklov. (POMI),
310(Kraev. Zadachi Mat. Fiz. i Smezh. Vopr. Tcor. Funkts. 35 [34]):191-212, 228-229,
2004.

[Syc05] M. A. Sychev. Scmicontinuity and relaxation theorems for integrands satisfying the fast

growth condition. Sibirsk. Mat. Zh., 46(3):679-697, 2005.
[You37] L. C. Young. Generalized curves and the existence of an attained absolute minimum in

the calculus of variations. Comptes Rendus de la Societe des Sciences et des Lettres de

Varsovie, 30:212-234, 1937.
[Zic89] William P. Ziemer. Weakly differentiate functions, volume 120 of Graduate Texts in

Mathematics. Springer- Verlag, New York, 1989. Sobolev spaces and functions of bounded

variation.

Laboratoire MIPA (Mathematiques, Informatique, Physique et Applications)
UNIVERSITE DE NIMES, Site des Carmes, Place Gabriel Peri, 30021 Nimes, France.
E-mail address: jean-philippe.mandallena@unimes.fr